Arch: resnet50_pt
Bs trn: 256
Bs val: 256
Hidden dim: 256
Dataset: celebA
Resample class: 
Slice with: rep
Rep cluster method: gmm
Num anchor: 32
Num positive: 32
Num negative: 32
Num negative easy: 0
Weight anc by loss: False
Weight pos by loss: False
Weight neg by loss: False
Anc loss temp: 0.5
Pos loss temp: 0.5
Neg loss temp: 0.5
Data wide pos: False
Target sample ratio: 1
Balance targets: False
Additional negatives: False
Hard negative factor: 0
Full contrastive: False
Train encoder: False
No projection head: False
Projection dim: 128
Batch factor: None
Temperature: 0.05
Single pos: False
Supervised linear scale up: False
Supervised update delay: 0
Contrastive weight: 0.5
Classifier update interval: 8
Optim: sgd
Max epoch: 50
Lr: 1e-05
Momentum: 0.9
Weight decay: 0.1
Weight decay c: 0.1
Stopping window: 30
Load encoder: 
Freeze encoder: False
Finetune epochs: 0
Clip grad norm: False
Lr scheduler classifier: 
Lr scheduler: 
Grad clip grad norm: False
Erm: False
Erm only: False
Pretrained spurious path: ./model/celebA/stage_one_erm_model_b_worst_avg_gap_best_seed0.pt
Max epoch s: 1
Bs trn s: 32
Lr s: 0.001
Momentum s: 0.9
Weight decay s: 0.0005
Slice temp: 10
Log loss interval: 10
Checkpoint interval: 50
Grad checkpoint interval: 50
Log visual interval: 100
Log grad visual interval: 50
Verbose: True
Seed: 0
Replicate: 0
No cuda: False
Resume: False
New slice: False
Num workers: 32
Evaluate: False
Data cmap: hsv
Test cmap: 
P correlation: 0.9
P corr by class: None
Train classes: ['blond', 'nonblond']
Train class ratios: None
Test shift: random
Flipped: False
Q: 0.7
Pretrained bmodel: True
Cosine: False
Exp: None
Supervised contrast: True
Prioritize spurious pos: False
Contrastive type: cnc
Compute auroc: False
Model type: resnet50_pt_cnc
Criterion: cross_entropy
Pretrained: False
Max grad norm: 1.0
Adam epsilon: 1e-08
Warmup steps: 0
Max grad norm s: 1.0
Adam epsilon s: 1e-08
Warmup steps s: 0
Grad max grad norm: 1.0
Grad adam epsilon: 1e-08
Grad warmup steps: 0
Device: cuda
Img file type: .png
Display image: False
Image path: ./images/celebA/celebA/config/contrastive_umaps
Log interval: 1
Log path: ./logs/celebA/config
Results path: ./results/celebA/config
Model path: ./model/celebA/config
Loss factor: 1
Supersample labels: False
Subsample labels: False
Weigh slice samples by loss: True
Val split: 0.2
Spurious train split: 0.2
Subsample groups: False
Train method: sc
Max robust acc: -1
Max robust epoch: -1
Max robust group acc: (None, None)
Root dir: ./datasets/data/CelebA/
Target name: Blond_Hair
Confounder names: ['Male']
Image mean: 0.449
Image std: 0.226
Augment data: False
Task: celebA
Num classes: 2
Experiment configs: config
Experiment name: cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0
Mi resampled: None

Loading checkpoints for train split:
[-1 -1 -1 ... -1 -1 -1]
<class 'numpy.ndarray'>
[0 1 2 3] [71629 66874 22880  1387]
Loading checkpoints for val split:
[-1 -1 -1 ... -1  1 -1]
<class 'numpy.ndarray'>
[0 1 2 3] [8535 8276 2874  182]
Loading checkpoints for test split:
[-1 -1 -1 ... -1 -1  1]
<class 'numpy.ndarray'>
[0 1 2 3] [9767 7535 2480  180]
Train dataset:
    Blond_Hair = 0, Male = 0 : n = 71629
    Blond_Hair = 0, Male = 1 : n = 66874
    Blond_Hair = 1, Male = 0 : n = 22880
    Blond_Hair = 1, Male = 1 : n = 1387
Val dataset:
    Blond_Hair = 0, Male = 0 : n = 8535
    Blond_Hair = 0, Male = 1 : n = 8276
    Blond_Hair = 1, Male = 0 : n = 2874
    Blond_Hair = 1, Male = 1 : n = 182
Test dataset:
    Blond_Hair = 0, Male = 0 : n = 9767
    Blond_Hair = 0, Male = 1 : n = 7535
    Blond_Hair = 1, Male = 0 : n = 2480
    Blond_Hair = 1, Male = 1 : n = 180
------------------------
> Loading spurious model
------------------------
Pretrained model loaded from ./model/celebA/stage_one_erm_model_b_worst_avg_gap_best_seed0.pt
======
# Calculate probability ...
======
======
# Load biased model ...
======
Epoch:   1 | Train Loss: 0.003 | Train Acc: 65.022 | Val Loss: 0.002 | Val Acc: 83.234
Training:
Accuracies by groups:
0, 0  acc: 22927 / 41612 =  55.097
0, 1  acc: 10751 / 14651 =  73.381
1, 0  acc: 67007 / 95317 =  70.299
1, 1  acc:  5152 / 11190 =  46.041
--------------------------------------
Average acc: 105837 / 162770 =  65.022
Robust  acc:  5152 / 11190 =  46.041
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6557 /  8535 =  76.825
0, 1  acc:  7094 /  8276 =  85.718
1, 0  acc:  2749 /  2874 =  95.651
1, 1  acc:   136 /   182 =  74.725
------------------------------------
Average acc: 16536 / 19867 =  83.234
Robust  acc:   136 /   182 =  74.725
------------------------------------------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 86.660
Robust Acc: 0.000 | Best Acc: 99.990
------------------------------------
Training, Epoch 0:
Accuracies by groups:
0, 0  acc:  9766 /  9767 =  99.990
0, 1  acc:  7533 /  7535 =  99.973
1, 0  acc:     0 /  2480 =   0.000
1, 1  acc:     0 /   180 =   0.000
-------------------------------------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 84.946
Robust Acc: 76.111 | Best Acc: 95.685
-------------------------------------
Training, Epoch 0:
Accuracies by groups:
0, 0  acc:  7981 /  9767 =  81.714
0, 1  acc:  6466 /  7535 =  85.813
1, 0  acc:  2373 /  2480 =  95.685
1, 1  acc:   137 /   180 =  76.111
------------------------------------
Average acc: 16957 / 19962 =  84.946
Robust  acc:   137 /   180 =  76.111
------------------------------------
Accuracies by groups:
0, 0  acc:  7981 /  9767 =  81.714
0, 1  acc:  6466 /  7535 =  85.813
1, 0  acc:  2373 /  2480 =  95.685
1, 1  acc:   137 /   180 =  76.111
------------------------------------
Average acc: 16957 / 19962 =  84.946
Robust  acc:   137Epoch:   2 | Train Loss: 0.005 | Train Acc: 16.859 | Val Loss: 0.002 | Val Acc: 84.608
Training:
Accuracies by groups:
0, 0  acc: 22040 / 22040 = 100.000
0, 1  acc:  5402 /  5402 = 100.000
1, 0  acc:     0 / 116186 =   0.000
1, 1  acc:     0 / 19142 =   0.000
-------------------------------------
Average acc: 27442 / 162770 =  16.859
Robust  acc:     0 / 116186 =   0.000
-Epoch:   2 | Train Loss: 0.002 | Train Acc: 74.819 | Val Loss: 0.002 | Val Acc: 81.789
Training:
Accuracies by groups:
0, 0  acc: 18363 / 41620 =  44.121
0, 1  acc:  7922 / 14544 =  54.469
1, 0  acc: 86838 / 95209 =  91.208
1, 1  acc:  8660 / 11397 =  75.985
--------------------------------------
Average acc: 121783 / 162770 =  74.819
Robust  acc: 18363 / 41620 =  44.121
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6578 /  8535 =  77.071
0, 1  acc:  6757 /  8276 =  81.646
1, 0  acc:  2768 /  2874 =  96.312
1, 1  acc:   146 /   182 =  80.220
------------------------------------
Average acc: 16249 / 19867 =  81.789
Robust  acc:  6578 /  8535 =  77.071
------------------------------------
New max robust acc: 77.07088459285296
debias model - Saving best checkpoint at epoch 1
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=0-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=1-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 83.383
Robust Acc: 80.556 | Best Acc: 96Epoch:   3 | Train Loss: 0.004 | Train Acc: 16.746 | Val Loss: 0.002 | Val Acc: 83.400
Training:
Accuracies by groups:
0, 0  acc: 21793 / 21795 =  99.991
0, 1  acc:  5365 /  5376 =  99.795
1, 0  acc:    82 / 116557 =   0.070
1, 1  acc:    18 / 19042 =   0.095
-------------------------------------
Average acc: 27258 / 162770 =  16.746
Robust  acc:    82 / 116557 =   0.070
-------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  8497 /  8535 =  99.555
0, 1  acc:  8043 /  8276 =  97.185
1, 0  acc:    23 /  2874 =   0.800
1, 1  acc:     6 /   182 =   3.297
------------------------------------
Average acc: 16569 / 19867 =  83.400
Robust  acc:    23 /  2874 =   0.800
------------------------------------
New max robust acc: 0.8002783576896312
debias model - Saving best checkpoint at epoch 2
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0Epoch:   3 | Train Loss: 0.002 | Train Acc: 76.437 | Val Loss: 0.002 | Val Acc: 83.254
Training:
Accuracies by groups:
0, 0  acc: 20205 / 41831 =  48.301
0, 1  acc:  7635 / 14397 =  53.032
1, 0  acc: 87611 / 95370 =  91.864
1, 1  acc:  8965 / 11172 =  80.245
--------------------------------------
Average acc: 124416 / 162770 =  76.437
Robust  acc: 20205 / 41831 =  48.301
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6814 /  8535 =  79.836
0, 1  acc:  6842 /  8276 =  82.673
1, 0  acc:  2737 /  2874 =  95.233
1, 1  acc:   147 /   182 =  80.769
------------------------------------
Average acc: 16540 / 19867 =  83.254
Robust  acc:  6814 /  8535 =  79.836
------------------------------------
New max robust acc: 79.83596953719977
debias model - Saving best checkpoint at epoch 2
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=2-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=2-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 84.636
Robust Acc: 81.111 | Best Acc: Epoch:   4 | Train Loss: 0.003 | Train Acc: 42.563 | Val Loss: 0.003 | Val Acc: 62.737
Training:
Accuracies by groups:
0, 0  acc: 18276 / 21649 =  84.420
0, 1  acc:  4037 /  5455 =  74.005
1, 0  acc: 36968 / 116539 =  31.722
1, 1  acc:  9998 / 19127 =  52.272
-------------------------------------
Average acc: 69279 / 162770 =  42.563
Robust  acc: 36968 / 116539 =  31.722
-------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6030 /  8535 =  70.650
0, 1  acc:  4190 /  8276 =  50.628
1, 0  acc:  2075 /  2874 =  72.199
1, 1  acc:   169 /   182 =  92.857
------------------------------------
Average acc: 12464 / 19867 =  62.737
Robust  acc:  4190 /  8276 =  50.628
------------------------------------
New max robust acc: 50.628322861285646
debias model - Saving best checkpoint at epoch 3
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1Epoch:   4 | Train Loss: 0.002 | Train Acc: 77.639 | Val Loss: 0.002 | Val Acc: 83.475
Training:
Accuracies by groups:
0, 0  acc: 21690 / 41925 =  51.735
0, 1  acc:  7734 / 14554 =  53.140
1, 0  acc: 87722 / 95056 =  92.285
1, 1  acc:  9227 / 11235 =  82.127
--------------------------------------
Average acc: 126373 / 162770 =  77.639
Robust  acc: 21690 / 41925 =  51.735
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6831 /  8535 =  80.035
0, 1  acc:  6854 /  8276 =  82.818
1, 0  acc:  2745 /  2874 =  95.511
1, 1  acc:   154 /   182 =  84.615
------------------------------------
Average acc: 16584 / 19867 =  83.475
Robust  acc:  6831 /  8535 =  80.035
------------------------------------
New max robust acc: 80.03514938488576
debias model - Saving best checkpoint at epoch 3
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=3-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=3-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 84.841
Robust Acc: 82.814 | Best Acc: Epoch:   5 | Train Loss: 0.002 | Train Acc: 79.102 | Val Loss: 0.003 | Val Acc: 45.326
Training:
Accuracies by groups:
0, 0  acc:  5901 / 22006 =  26.815
0, 1  acc:  1848 /  5513 =  33.521
1, 0  acc: 102637 / 116302 =  88.250
1, 1  acc: 18368 / 18949 =  96.934
--------------------------------------
Average acc: 128754 / 162770 =  79.102
Robust  acc:  5901 / 22006 =  26.815
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3772 /  8535 =  44.194
0, 1  acc:  2230 /  8276 =  26.945
1, 0  acc:  2822 /  2874 =  98.191
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc:  9005 / 19867 =  45.326
Robust  acc:  2230 /  8276 =  26.945
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 45.511
Robust Acc: 26.238 | Best Acc: 98.333
-------------------------------------
Training, Epoch 4:
Accuracies by groups:
0, 0  acc:  4504 /  9767 =  46.114
0, 1  acc:  1977 /  7535 =Epoch:   5 | Train Loss: 0.002 | Train Acc: 78.706 | Val Loss: 0.002 | Val Acc: 83.369
Training:
Accuracies by groups:
0, 0  acc: 22406 / 41667 =  53.774
0, 1  acc:  8007 / 14351 =  55.794
1, 0  acc: 88307 / 95457 =  92.510
1, 1  acc:  9389 / 11295 =  83.125
--------------------------------------
Average acc: 128109 / 162770 =  78.706
Robust  acc: 22406 / 41667 =  53.774
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6811 /  8535 =  79.801
0, 1  acc:  6839 /  8276 =  82.637
1, 0  acc:  2755 /  2874 =  95.859
1, 1  acc:   158 /   182 =  86.813
------------------------------------
Average acc: 16563 / 19867 =  83.369
Robust  acc:  6811 /  8535 =  79.801
------------------------------------
-------------------------------------------
Avg Test Loss: 0.001 | Avg Test Acc: 84.596
Robust Acc: 82.455 | Best Acc: 95Epoch:   6 | Train Loss: 0.002 | Train Acc: 83.622 | Val Loss: 0.003 | Val Acc: 41.466
Training:
Accuracies by groups:
0, 0  acc:  2663 / 21868 =  12.178
0, 1  acc:  1218 /  5412 =  22.506
1, 0  acc: 113025 / 116158 =  97.303
1, 1  acc: 19205 / 19332 =  99.343
--------------------------------------
Average acc: 136111 / 162770 =  83.622
Robust  acc:  2663 / 21868 =  12.178
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3308 /  8535 =  38.758
0, 1  acc:  1894 /  8276 =  22.885
1, 0  acc:  2855 /  2874 =  99.339
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc:  8238 / 19867 =  41.466
Robust  acc:  1894 /  8276 =  22.885
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 41.875
Robust Acc: 22.163 | Best Acc: 98.871
-------------------------------------
Training, Epoch 5:
Accuracies by groups:
0, 0  acc:  4060 /  9767 =  41.569
0, 1  acc:  1670 /  7535Epoch:   6 | Train Loss: 0.002 | Train Acc: 79.212 | Val Loss: 0.001 | Val Acc: 84.270
Training:
Accuracies by groups:
0, 0  acc: 23393 / 42007 =  55.688
0, 1  acc:  8208 / 14519 =  56.533
1, 0  acc: 87922 / 94900 =  92.647
1, 1  acc:  9411 / 11344 =  82.960
--------------------------------------
Average acc: 128934 / 162770 =  79.212
Robust  acc: 23393 / 42007 =  55.688
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  6880 /  8535 =  80.609
0, 1  acc:  6956 /  8276 =  84.050
1, 0  acc:  2748 /  2874 =  95.616
1, 1  acc:   158 /   182 =  86.813
------------------------------------
Average acc: 16742 / 19867 =  84.270
Robust  acc:  6880 /  8535 =  80.609
------------------------------------
New max robust acc: 80.60925600468659
debias model - Saving best checkpoint at epoch 5
replace: True
-> Updating checkpointEpoch:   7 | Train Loss: 0.002 | Train Acc: 84.130 | Val Loss: 0.003 | Val Acc: 41.828
Training:
Accuracies by groups:
0, 0  acc:  2433 / 21802 =  11.160
0, 1  acc:  1163 /  5579 =  20.846
1, 0  acc: 114252 / 116201 =  98.323
1, 1  acc: 19091 / 19188 =  99.494
--------------------------------------
Average acc: 136939 / 162770 =  84.130
Robust  acc:  2433 / 21802 =  11.160
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3288 /  8535 =  38.524
0, 1  acc:  1982 /  8276 =  23.949
1, 0  acc:  2859 /  2874 =  99.478
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc:  8310 / 19867 =  41.828
Robust  acc:  1982 /  8276 =  23.949
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 42.586
Robust Acc: 22.880 | Best Acc: 99.444
-------------------------------------
Training, Epoch 6:
Accuracies by groups:
0, 0  acc:  4139 /  9767 =  42.377
0, 1  acc:  1724 /  7535 =  22.880
1, 0  acc:  2459 /  2480 =  99.153
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  8501 / 19962 =  42.586
Robust  acc:  1724 /  7535 =  22.880
------------------------------------
Accuracies by groups:
0, 0  acc:  4139 /  9767 =  42.377
0, 1  acc:  1724 /  7535 =  22.880
1, 0  acc:  2459 /  2480 =  99.153
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  8501 / 19962 =  42.586
Robust  acc:  1724 /  7535 =  22.880
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  4139 /  9767 =  42.377
0, 1  acc:  1724 /Epoch:   7 | Train Loss: 0.002 | Train Acc: 79.982 | Val Loss: 0.001 | Val Acc: 84.361
Training:
Accuracies by groups:
0, 0  acc: 23725 / 41726 =  56.859
0, 1  acc:  8536 / 14445 =  59.093
1, 0  acc: 88682 / 95408 =  92.950
1, 1  acc: Epoch:   8 | Train Loss: 0.002 | Train Acc: 84.533 | Val Loss: 0.003 | Val Acc: 43.867
Training:
Accuracies by groups:
0, 0  acc:  2470 / 21674 =  11.396
0, 1  acc:  1308 /  5436 =  24.062
1, 0  acc: 114913 / 116666 =  98.497
1, 1  acc: 18904 / 18994 =  99.526
--------------------------------------
Average acc: 137595 / 162770 =  84.533
Robust  acc:  2470 / 21674 =  11.396
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3428 /  8535 =  40.164
0, 1  acc:  2246 /  8276 =  27.139
1, 0  acc:  2861 /  2874 =  99.548
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc:  8715 / 19867 =  43.867
Robust  acc:  2246 /  8276 =  27.139
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 44.840
Robust Acc: 25.985 | Best Acc: 99.444
-------------------------------------
Training, Epoch 7:
Accuracies by groups:
0, 0  acc:  4353 /  9767 =  44.568
0, 1  acc:  1958 /  7535 =  25.985
1, 0  acc:  2461 /  2480 =  99.234
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  8951 / 19962 =  44.840
Robust  acc:  1958 /  7535 =  25.985
------------------------------------
Accuracies by groups:
0, 0  acc:  4353 /  9767 =  44.568
0, 1  acc:  1958 /  7535 =  25.985
1, 0  acc:  2461 /  2480 =  99.234
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  8951 / 19962 =  44.840
Robust  acc:  1958 /  7535 =  25.985
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  4353 /  9767 =  44.568
0, 1  acc:  1958Epoch:   8 | Train Loss: 0.002 | Train Acc: 80.316 | Val Loss: 0.001 | Val Acc: 84.895
Training:
Accuracies by groups:
0, 0  acc: 24270 / 42020 =  57.758
0, 1  acc:  8928 / 14638 =  60.992
1, 0  acc: 88081 / 94691 =  93.019
1, 1  acc:  9Epoch:   9 | Train Loss: 0.002 | Train Acc: 84.793 | Val Loss: 0.003 | Val Acc: 52.570
Training:
Accuracies by groups:
0, 0  acc:  3062 / 21686 =  14.120
0, 1  acc:  1580 /  5510 =  28.675
1, 0  acc: 114349 / 116423 =  98.219
1, 1  acc: 19027 / 19151 =  99.353
--------------------------------------
Average acc: 138018 / 162770 =  84.793
Robust  acc:  3062 / 21686 =  14.120
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4232 /  8535 =  49.584
0, 1  acc:  3234 /  8276 =  39.077
1, 0  acc:  2798 /  2874 =  97.356
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 10444 / 19867 =  52.570
Robust  acc:  3234 /  8276 =  39.077
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 53.857
Robust Acc: 37.983 | Best Acc: 99.444
-------------------------------------
Training, Epoch 8:
Accuracies by groups:
0, 0  acc:  5312 /  9767 =  54.387
0, 1  acc:  2862 /  7535 =  37.983
1, 0  acc:  2398 /  2480 =  96.694
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10751 / 19962 =  53.857
Robust  acc:  2862 /  7535 =  37.983
------------------------------------
Accuracies by groups:
0, 0  acc:  5312 /  9767 =  54.387
0, 1  acc:  2862 /  7535 =  37.983
1, 0  acc:  2398 /  2480 =  96.694
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10751 / 19962 =  53.857
Robust  acc:  2862 /  7535 =  37.983
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5312 /  9767 =  54.387
0, 1  acc:  2862 /  7535 =  37.983
1, 0  acc:  2398 /  2480 =  96.694
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10751 / 19962 =  53.857
Robust  acc:  2862 /  7535 =  37.983
------------------------------------
Epoch:  10 | Train Loss: 0.002 | Train Acc: 84.963 | Val Loss: 0.003 | Val Acc: 46.016
Training:
Accuracies by groups:
0, 0  acc:  3949 / 21821 =  18.097
0, 1  acc:  1789 /  5514 =  32.445
1, 0  acc: 113353 / 116123 =  97.615
1, 1  acc: 19204 / 19312 =  99.441
--------------------------------------
Average acc: 138295 / 162770 =  84.963
Robust  acc:  3949 / 21821 =  18.097
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3633 /  8535 =  42.566
0, 1  acc:  2637 /  8276 =  31.863
1, 0  acc:  2691 /  2874 =  93.633
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc:  9142 / 19867 =  46.016
Robust  acc:  2637 /  8276 =  31.863
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 46.749
Robust Acc: 29.993 | Best Acc: 99.444
-------------------------------------
Training, Epoch 9:
Accuracies by groups:
0, 0  acc:  4576 /  9767 =  46.852
0, 1  acc:  2260 /  7535 =  29.993
1, 0  acc:  2317 /  2480 =  93.427
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  9332 / 19962 =  46.749
Robust  acc:  2260 /  7535 =  29.993
------------------------------------
Accuracies by groups:
0, 0  acc:  4576 /  9767 =  46.852
0, 1  acc:  2260 /  7535 =  29.993
1, 0  acc:  2317 /  2480 =  93.427
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  9332 / 19962 =  46.749
Robust  acc:  2260 /  7535 =  29.993
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  4576 /  9767 =  46.852
0, 1  acc:  2260 /  7535 =  29.993
1, 0  acc:  2317 /  2480 =  93.427
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc:  9332 / 19962 =  46.749
Robust  acc:  2260 /  7535 =  29.993
------------------------------------
Epoch:  11 | Train Loss: 0.002 | Train Acc: 85.167 | Val Loss: 0.003 | Val Acc: 50.491
Training:
Accuracies by groups:
0, 0  acc:  4474 / 21849 =  20.477
0, 1  acc:  1917 /  5389 =  35.572
1, 0  acc: 113063 / 116237 =  97.269
1, 1  acc: 19173 / 19295 =  99.368
--------------------------------------
Average acc: 138627 / 162770 =  85.167
Robust  acc:  4474 / 21849 =  20.477
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4011 /  8535 =  46.995
0, 1  acc:  3296 /  8276 =  39.826
1, 0  acc:  2543 /  2874 =  88.483
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10031 / 19867 =  50.491
Robust  acc:  3296 /  8276 =  39.826
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 51.854
Robust Acc: 38.381 | Best Acc: 98.889
-------------------------------------
Training, Epoch 10:
Accuracies by groups:
0, 0  acc:  5082 /  9767 =  52.032
0, 1  acc:  2892 /  7535 =  38.381
1, 0  acc:  2199 /  2480 =  88.669
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10351 / 19962 =  51.854
Robust  acc:  2892 /  7535 =  38.381
------------------------------------
Accuracies by groups:
0, 0  acc:  5082 /  9767 =  52.032
0, 1  acc:  2892 /  7535 =  38.381
1, 0  acc:  2199 /  2480 =  88.669
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10351 / 19962 =  51.854
Robust  acc:  2892 /  7535 =  38.381
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5082 /  9767 =  52.032
0, 1  acc:  2892 /  7535 =  38.381
1, 0  acc:  2199 /  2480 =  88.669
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10351 / 19962 =  51.854
Robust  acc:  2892 /  7535 =  38.381
------------------------------------
Epoch:  12 | Train Loss: 0.002 | Train Acc: 85.374 | Val Loss: 0.003 | Val Acc: 49.444
Training:
Accuracies by groups:
0, 0  acc:  5035 / 21839 =  23.055
0, 1  acc:  2070 /  5476 =  37.801
1, 0  acc: 112792 / 116257 =  97.020
1, 1  acc: 19067 / 19198 =  99.318
--------------------------------------
Average acc: 138964 / 162770 =  85.374
Robust  acc:  5035 / 21839 =  23.055
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3919 /  8535 =  45.917
0, 1  acc:  3240 /  8276 =  39.149
1, 0  acc:  2483 /  2874 =  86.395
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc:  9823 / 19867 =  49.444
Robust  acc:  3240 /  8276 =  39.149
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 50.566
Robust Acc: 37.200 | Best Acc: 98.889
-------------------------------------
Training, Epoch 11:
Accuracies by groups:
0, 0  acc:  4943 /  9767 =  50.609
0, 1  acc:  2803 /  7535 =  37.200
1, 0  acc:  2170 /  2480 =  87.500
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10094 / 19962 =  50.566
Robust  acc:  2803 /  7535 =  37.200
------------------------------------
Accuracies by groups:
0, 0  acc:  4943 /  9767 =  50.609
0, 1  acc:  2803 /  7535 =  37.200
1, 0  acc:  2170 /  2480 =  87.500
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10094 / 19962 =  50.566
Robust  acc:  2803 /  7535 =  37.200
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  4943 /  9767 =  50.609
0, 1  acc:  2803 /  7535 =  37.200
1, 0  acc:  2170 /  2480 =  87.500
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10094 / 19962 =  50.566
Robust  acc:  2803 /  7535 =  37.200
------------------------------------
Epoch:  13 | Train Loss: 0.002 | Train Acc: 85.798 | Val Loss: 0.003 | Val Acc: 52.680
Training:
Accuracies by groups:
0, 0  acc:  5587 / 21575 =  25.896
0, 1  acc:  2140 /  5496 =  38.937
1, 0  acc: 112838 / 116446 =  96.902
1, 1  acc: 19088 / 19253 =  99.143
--------------------------------------
Average acc: 139653 / 162770 =  85.798
Robust  acc:  5587 / 21575 =  25.896
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4155 /  8535 =  48.682
0, 1  acc:  3625 /  8276 =  43.801
1, 0  acc:  2505 /  2874 =  87.161
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10466 / 19867 =  52.680
Robust  acc:  3625 /  8276 =  43.801
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 54.298
Robust Acc: 42.614 | Best Acc: 99.444
-------------------------------------
Training, Epoch 12:
Accuracies by groups:
0, 0  acc:  5263 /  9767 =  53.886
0, 1  acc:  3211 /  7535 =  42.614
1, 0  acc:  2186 /  2480 =  88.145
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10839 / 19962 =  54.298
Robust  acc:  3211 /  7535 =  42.614
------------------------------------
Accuracies by groups:
0, 0  acc:  5263 /  9767 =  53.886
0, 1  acc:  3211 /  7535 =  42.614
1, 0  acc:  2186 /  2480 =  88.145
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10839 / 19962 =  54.298
Robust  acc:  3211 /  7535 =  42.614
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5263 /  9767 =  53.886
0, 1  acc:  3211 /  7535 =  42.614
1, 0  acc:  2186 /  2480 =  88.145
1, 1  acc:   179 /   180 =  99.444
------------------------------------
Average acc: 10839 / 19962 =  54.298
Robust  acc:  3211 /  7535 =  42.614
------------------------------------
Epoch:  14 | Train Loss: 0.002 | Train Acc: 85.764 | Val Loss: 0.003 | Val Acc: 51.956
Training:
Accuracies by groups:
0, 0  acc:  6157 / 21893 =  28.123
0, 1  acc:  2284 /  5523 =  41.354
1, 0  acc: 112074 / 116115 =  96.520
1, 1  acc: 19083 / 19239 =  99.189
--------------------------------------
Average acc: 139598 / 162770 =  85.764
Robust  acc:  6157 / 21893 =  28.123
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4113 /  8535 =  48.190
0, 1  acc:  3583 /  8276 =  43.294
1, 0  acc:  2445 /  2874 =  85.073
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10322 / 19867 =  51.956
Robust  acc:  3583 /  8276 =  43.294
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 53.371
Robust Acc: 42.004 | Best Acc: 98.333
-------------------------------------
Training, Epoch 13:
Accuracies by groups:
0, 0  acc:  5198 /  9767 =  53.220
0, 1  acc:  3165 /  7535 =  42.004
1, 0  acc:  2114 /  2480 =  85.242
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10654 / 19962 =  53.371
Robust  acc:  3165 /  7535 =  42.004
------------------------------------
Accuracies by groups:
0, 0  acc:  5198 /  9767 =  53.220
0, 1  acc:  3165 /  7535 =  42.004
1, 0  acc:  2114 /  2480 =  85.242
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10654 / 19962 =  53.371
Robust  acc:  3165 /  7535 =  42.004
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5198 /  9767 =  53.220
0, 1  acc:  3165 /  7535 =  42.004
1, 0  acc:  2114 /  2480 =  85.242
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10654 / 19962 =  53.371
Robust  acc:  3165 /  7535 =  42.004
------------------------------------
Epoch:  15 | Train Loss: 0.001 | Train Acc: 85.995 | Val Loss: 0.003 | Val Acc: 54.900
Training:
Accuracies by groups:
0, 0  acc:  6632 / 21843 =  30.362
0, 1  acc:  2335 /  5530 =  42.224
1, 0  acc: 112227 / 116442 =  96.380
1, 1  acc: 18780 / 18955 =  99.077
--------------------------------------
Average acc: 139974 / 162770 =  85.995
Robust  acc:  6632 / 21843 =  30.362
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4372 /  8535 =  51.224
0, 1  acc:  3895 /  8276 =  47.064
1, 0  acc:  2459 /  2874 =  85.560
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10907 / 19867 =  54.900
Robust  acc:  3895 /  8276 =  47.064
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 56.462
Robust Acc: 46.264 | Best Acc: 98.889
-------------------------------------
Training, Epoch 14:
Accuracies by groups:
0, 0  acc:  5481 /  9767 =  56.118
0, 1  acc:  3486 /  7535 =  46.264
1, 0  acc:  2126 /  2480 =  85.726
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11271 / 19962 =  56.462
Robust  acc:  3486 /  7535 =  46.264
------------------------------------
Accuracies by groups:
0, 0  acc:  5481 /  9767 =  56.118
0, 1  acc:  3486 /  7535 =  46.264
1, 0  acc:  2126 /  2480 =  85.726
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11271 / 19962 =  56.462
Robust  acc:  3486 /  7535 =  46.264
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5481 /  9767 =  56.118
0, 1  acc:  3486 /  7535 =  46.264
1, 0  acc:  2126 /  2480 =  85.726
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11271 / 19962 =  56.462
Robust  acc:  3486 /  7535 =  46.264
------------------------------------
Epoch:  16 | Train Loss: 0.001 | Train Acc: 86.309 | Val Loss: 0.003 | Val Acc: 52.570
Training:
Accuracies by groups:
0, 0  acc:  7176 / 21812 =  32.899
0, 1  acc:  2448 /  5492 =  44.574
1, 0  acc: 111904 / 116338 =  96.189
1, 1  acc: 18957 / 19128 =  99.106
--------------------------------------
Average acc: 140485 / 162770 =  86.309
Robust  acc:  7176 / 21812 =  32.899
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4231 /  8535 =  49.572
0, 1  acc:  3758 /  8276 =  45.408
1, 0  acc:  2274 /  2874 =  79.123
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10444 / 19867 =  52.570
Robust  acc:  3758 /  8276 =  45.408
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 54.293
Robust Acc: 44.711 | Best Acc: 98.333
-------------------------------------
Training, Epoch 15:
Accuracies by groups:
0, 0  acc:  5309 /  9767 =  54.357
0, 1  acc:  3369 /  7535 =  44.711
1, 0  acc:  1983 /  2480 =  79.960
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10838 / 19962 =  54.293
Robust  acc:  3369 /  7535 =  44.711
------------------------------------
Accuracies by groups:
0, 0  acc:  5309 /  9767 =  54.357
0, 1  acc:  3369 /  7535 =  44.711
1, 0  acc:  1983 /  2480 =  79.960
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10838 / 19962 =  54.293
Robust  acc:  3369 /  7535 =  44.711
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5309 /  9767 =  54.357
0, 1  acc:  3369 /  7535 =  44.711
1, 0  acc:  1983 /  2480 =  79.960
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10838 / 19962 =  54.293
Robust  acc:  3369 /  7535 =  44.711
------------------------------------
Epoch:  17 | Train Loss: 0.001 | Train Acc: 86.631 | Val Loss: 0.003 | Val Acc: 52.071
Training:
Accuracies by groups:
0, 0  acc:  7844 / 21777 =  36.020
0, 1  acc:  2467 /  5392 =  45.753
1, 0  acc: 111732 / 116453 =  95.946
1, 1  acc: 18966 / 19148 =  99.050
--------------------------------------
Average acc: 141009 / 162770 =  86.631
Robust  acc:  7844 / 21777 =  36.020
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4121 /  8535 =  48.284
0, 1  acc:  3781 /  8276 =  45.686
1, 0  acc:  2262 /  2874 =  78.706
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10345 / 19867 =  52.071
Robust  acc:  3781 /  8276 =  45.686
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 53.662
Robust Acc: 44.990 | Best Acc: 98.333
-------------------------------------
Training, Epoch 16:
Accuracies by groups:
0, 0  acc:  5174 /  9767 =  52.974
0, 1  acc:  3390 /  7535 =  44.990
1, 0  acc:  1971 /  2480 =  79.476
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10712 / 19962 =  53.662
Robust  acc:  3390 /  7535 =  44.990
------------------------------------
Accuracies by groups:
0, 0  acc:  5174 /  9767 =  52.974
0, 1  acc:  3390 /  7535 =  44.990
1, 0  acc:  1971 /  2480 =  79.476
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10712 / 19962 =  53.662
Robust  acc:  3390 /  7535 =  44.990
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5174 /  9767 =  52.974
0, 1  acc:  3390 /  7535 =  44.990
1, 0  acc:  1971 /  2480 =  79.476
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 10712 / 19962 =  53.662
Robust  acc:  3390 /  7535 =  44.990
------------------------------------
Epoch:  18 | Train Loss: 0.001 | Train Acc: 86.583 | Val Loss: 0.003 | Val Acc: 55.127
Training:
Accuracies by groups:
0, 0  acc:  8056 / 21917 =  36.757
0, 1  acc:  2464 /  5418 =  45.478
1, 0  acc: 111555 / 116379 =  95.855
1, 1  acc: 18856 / 19056 =  98.950
--------------------------------------
Average acc: 140931 / 162770 =  86.583
Robust  acc:  8056 / 21917 =  36.757
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4311 /  8535 =  50.510
0, 1  acc:  4038 /  8276 =  48.792
1, 0  acc:  2423 /  2874 =  84.308
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 10952 / 19867 =  55.127
Robust  acc:  4038 /  8276 =  48.792
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 56.743
Robust Acc: 48.228 | Best Acc: 98.889
-------------------------------------
Training, Epoch 17:
Accuracies by groups:
0, 0  acc:  5432 /  9767 =  55.616
0, 1  acc:  3634 /  7535 =  48.228
1, 0  acc:  2083 /  2480 =  83.992
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11327 / 19962 =  56.743
Robust  acc:  3634 /  7535 =  48.228
------------------------------------
Accuracies by groups:
0, 0  acc:  5432 /  9767 =  55.616
0, 1  acc:  3634 /  7535 =  48.228
1, 0  acc:  2083 /  2480 =  83.992
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11327 / 19962 =  56.743
Robust  acc:  3634 /  7535 =  48.228
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5432 /  9767 =  55.616
0, 1  acc:  3634 /  7535 =  48.228
1, 0  acc:  2083 /  2480 =  83.992
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11327 / 19962 =  56.743
Robust  acc:  3634 /  7535 =  48.228
------------------------------------
Epoch:  19 | Train Loss: 0.001 | Train Acc: 86.953 | Val Loss: 0.003 | Val Acc: 53.561
Training:
Accuracies by groups:
0, 0  acc:  8600 / 21791 =  39.466
0, 1  acc:  2575 /  5452 =  47.230
1, 0  acc: 111381 / 116354 =  95.726
1, 1  acc: 18978 / 19173 =  98.983
--------------------------------------
Average acc: 141534 / 162770 =  86.953
Robust  acc:  8600 / 21791 =  39.466
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4168 /  8535 =  48.834
0, 1  acc:  3848 /  8276 =  46.496
1, 0  acc:  2444 /  2874 =  85.038
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10641 / 19867 =  53.561
Robust  acc:  3848 /  8276 =  46.496
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 54.814
Robust Acc: 45.362 | Best Acc: 98.889
-------------------------------------
Training, Epoch 18:
Accuracies by groups:
0, 0  acc:  5264 /  9767 =  53.896
0, 1  acc:  3418 /  7535 =  45.362
1, 0  acc:  2082 /  2480 =  83.952
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10942 / 19962 =  54.814
Robust  acc:  3418 /  7535 =  45.362
------------------------------------
Accuracies by groups:
0, 0  acc:  5264 /  9767 =  53.896
0, 1  acc:  3418 /  7535 =  45.362
1, 0  acc:  2082 /  2480 =  83.952
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10942 / 19962 =  54.814
Robust  acc:  3418 /  7535 =  45.362
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5264 /  9767 =  53.896
0, 1  acc:  3418 /  7535 =  45.362
1, 0  acc:  2082 /  2480 =  83.952
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10942 / 19962 =  54.814
Robust  acc:  3418 /  7535 =  45.362
------------------------------------
Epoch:  20 | Train Loss: 0.001 | Train Acc: 87.155 | Val Loss: 0.003 | Val Acc: 61.771
Training:
Accuracies by groups:
0, 0  acc:  8894 / 21673 =  41.037
0, 1  acc:  2627 /  5606 =  46.861
1, 0  acc: 111360 / 116331 =  95.727
1, 1  acc: 18981 / 19160 =  99.066
--------------------------------------
Average acc: 141862 / 162770 =  87.155
Robust  acc:  8894 / 21673 =  41.037
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4841 /  8535 =  56.719
0, 1  acc:  4652 /  8276 =  56.211
1, 0  acc:  2599 /  2874 =  90.431
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 12272 / 19867 =  61.771
Robust  acc:  4652 /  8276 =  56.211
------------------------------------
New max robust acc: 56.21072982116965
debias model - Saving best checkpoint at epoch 19
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=16-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=19-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 63.511
Robust Acc: 55.806 | Best Acc: 98.889
-------------------------------------
Training, Epoch 19:
Accuracies by groups:
0, 0  acc:  6054 /  9767 =  61.984
0, 1  acc:  4205 /  7535 =  55.806
1, 0  acc:  2241 /  2480 =  90.363
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12678 / 19962 =  63.511
Robust  acc:  4205 /  7535 =  55.806
------------------------------------
Accuracies by groups:
0, 0  acc:  6054 /  9767 =  61.984
0, 1  acc:  4205 /  7535 =  55.806
1, 0  acc:  2241 /  2480 =  90.363
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12678 / 19962 =  63.511
Robust  acc:  4205 /  7535 =  55.806
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6054 /  9767 =  61.984
0, 1  acc:  4205 /  7535 =  55.806
1, 0  acc:  2241 /  2480 =  90.363
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12678 / 19962 =  63.511
Robust  acc:  4205 /  7535 =  55.806
------------------------------------
Epoch:  21 | Train Loss: 0.001 | Train Acc: 87.452 | Val Loss: 0.003 | Val Acc: 56.093
Training:
Accuracies by groups:
0, 0  acc:  9652 / 21926 =  44.021
0, 1  acc:  2542 /  5308 =  47.890
1, 0  acc: 111203 / 116379 =  95.552
1, 1  acc: 18949 / 19157 =  98.914
--------------------------------------
Average acc: 142346 / 162770 =  87.452
Robust  acc:  9652 / 21926 =  44.021
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4408 /  8535 =  51.646
0, 1  acc:  4318 /  8276 =  52.175
1, 0  acc:  2238 /  2874 =  77.871
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 11144 / 19867 =  56.093
Robust  acc:  4408 /  8535 =  51.646
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 57.775
Robust Acc: 51.175 | Best Acc: 95.556
-------------------------------------
Training, Epoch 20:
Accuracies by groups:
0, 0  acc:  5579 /  9767 =  57.121
0, 1  acc:  3856 /  7535 =  51.175
1, 0  acc:  1926 /  2480 =  77.661
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 11533 / 19962 =  57.775
Robust  acc:  3856 /  7535 =  51.175
------------------------------------
Accuracies by groups:
0, 0  acc:  5579 /  9767 =  57.121
0, 1  acc:  3856 /  7535 =  51.175
1, 0  acc:  1926 /  2480 =  77.661
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 11533 / 19962 =  57.775
Robust  acc:  3856 /  7535 =  51.175
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5579 /  9767 =  57.121
0, 1  acc:  3856 /  7535 =  51.175
1, 0  acc:  1926 /  2480 =  77.661
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 11533 / 19962 =  57.775
Robust  acc:  3856 /  7535 =  51.175
------------------------------------
Epoch:  22 | Train Loss: 0.001 | Train Acc: 87.627 | Val Loss: 0.003 | Val Acc: 59.325
Training:
Accuracies by groups:
0, 0  acc: 10156 / 21848 =  46.485
0, 1  acc:  2624 /  5541 =  47.356
1, 0  acc: 111098 / 116430 =  95.420
1, 1  acc: 18753 / 18951 =  98.955
--------------------------------------
Average acc: 142631 / 162770 =  87.627
Robust  acc: 10156 / 21848 =  46.485
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4646 /  8535 =  54.435
0, 1  acc:  4467 /  8276 =  53.975
1, 0  acc:  2493 /  2874 =  86.743
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 11786 / 19867 =  59.325
Robust  acc:  4467 /  8276 =  53.975
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 60.565
Robust Acc: 52.940 | Best Acc: 98.889
-------------------------------------
Training, Epoch 21:
Accuracies by groups:
0, 0  acc:  5802 /  9767 =  59.404
0, 1  acc:  3989 /  7535 =  52.940
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12090 / 19962 =  60.565
Robust  acc:  3989 /  7535 =  52.940
------------------------------------
Accuracies by groups:
0, 0  acc:  5802 /  9767 =  59.404
0, 1  acc:  3989 /  7535 =  52.940
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12090 / 19962 =  60.565
Robust  acc:  3989 /  7535 =  52.940
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5802 /  9767 =  59.404
0, 1  acc:  3989 /  7535 =  52.940
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12090 / 19962 =  60.565
Robust  acc:  3989 /  7535 =  52.940
------------------------------------
Epoch:  23 | Train Loss: 0.001 | Train Acc: 88.038 | Val Loss: 0.003 | Val Acc: 55.227
Training:
Accuracies by groups:
0, 0  acc: 10577 / 21675 =  48.798
0, 1  acc:  2672 /  5414 =  49.354
1, 0  acc: 111051 / 116466 =  95.351
1, 1  acc: 19000 / 19215 =  98.881
--------------------------------------
Average acc: 143300 / 162770 =  88.038
Robust  acc: 10577 / 21675 =  48.798
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4121 /  8535 =  48.284
0, 1  acc:  4108 /  8276 =  49.638
1, 0  acc:  2562 /  2874 =  89.144
1, 1  acc:   181 /   182 =  99.451
------------------------------------
Average acc: 10972 / 19867 =  55.227
Robust  acc:  4121 /  8535 =  48.284
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 56.748
Robust Acc: 48.600 | Best Acc: 98.889
-------------------------------------
Training, Epoch 22:
Accuracies by groups:
0, 0  acc:  5284 /  9767 =  54.101
0, 1  acc:  3662 /  7535 =  48.600
1, 0  acc:  2204 /  2480 =  88.871
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11328 / 19962 =  56.748
Robust  acc:  3662 /  7535 =  48.600
------------------------------------
Accuracies by groups:
0, 0  acc:  5284 /  9767 =  54.101
0, 1  acc:  3662 /  7535 =  48.600
1, 0  acc:  2204 /  2480 =  88.871
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11328 / 19962 =  56.748
Robust  acc:  3662 /  7535 =  48.600
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5284 /  9767 =  54.101
0, 1  acc:  3662 /  7535 =  48.600
1, 0  acc:  2204 /  2480 =  88.871
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 11328 / 19962 =  56.748
Robust  acc:  3662 /  7535 =  48.600
------------------------------------
Epoch:  24 | Train Loss: 0.001 | Train Acc: 88.239 | Val Loss: 0.003 | Val Acc: 57.336
Training:
Accuracies by groups:
0, 0  acc: 11070 / 21817 =  50.740
0, 1  acc:  2705 /  5518 =  49.021
1, 0  acc: 110901 / 116291 =  95.365
1, 1  acc: 18950 / 19144 =  98.987
--------------------------------------
Average acc: 143626 / 162770 =  88.239
Robust  acc:  2705 /  5518 =  49.021
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4389 /  8535 =  51.424
0, 1  acc:  4415 /  8276 =  53.347
1, 0  acc:  2407 /  2874 =  83.751
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 11391 / 19867 =  57.336
Robust  acc:  4389 /  8535 =  51.424
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 58.892
Robust Acc: 52.064 | Best Acc: 98.333
-------------------------------------
Training, Epoch 23:
Accuracies by groups:
0, 0  acc:  5578 /  9767 =  57.111
0, 1  acc:  3923 /  7535 =  52.064
1, 0  acc:  2078 /  2480 =  83.790
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 11756 / 19962 =  58.892
Robust  acc:  3923 /  7535 =  52.064
------------------------------------
Accuracies by groups:
0, 0  acc:  5578 /  9767 =  57.111
0, 1  acc:  3923 /  7535 =  52.064
1, 0  acc:  2078 /  2480 =  83.790
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 11756 / 19962 =  58.892
Robust  acc:  3923 /  7535 =  52.064
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5578 /  9767 =  57.111
0, 1  acc:  3923 /  7535 =  52.064
1, 0  acc:  2078 /  2480 =  83.790
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 11756 / 19962 =  58.892
Robust  acc:  3923 /  7535 =  52.064
------------------------------------
Epoch:  25 | Train Loss: 0.001 | Train Acc: 88.803 | Val Loss: 0.003 | Val Acc: 62.536
Training:
Accuracies by groups:
0, 0  acc: 11862 / 21842 =  54.308
0, 1  acc:  2721 /  5476 =  49.690
1, 0  acc: 111174 / 116476 =  95.448
1, 1  acc: 18788 / 18976 =  99.009
--------------------------------------
Average acc: 144545 / 162770 =  88.803
Robust  acc:  2721 /  5476 =  49.690
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4918 /  8535 =  57.622
0, 1  acc:  4903 /  8276 =  59.244
1, 0  acc:  2425 /  2874 =  84.377
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 12424 / 19867 =  62.536
Robust  acc:  4918 /  8535 =  57.622
------------------------------------
New max robust acc: 57.6215582893966
debias model - Saving best checkpoint at epoch 24
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=23-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=24-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.493
Robust Acc: 58.872 | Best Acc: 96.667
-------------------------------------
Training, Epoch 24:
Accuracies by groups:
0, 0  acc:  6143 /  9767 =  62.895
0, 1  acc:  4436 /  7535 =  58.872
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12874 / 19962 =  64.493
Robust  acc:  4436 /  7535 =  58.872
------------------------------------
Accuracies by groups:
0, 0  acc:  6143 /  9767 =  62.895
0, 1  acc:  4436 /  7535 =  58.872
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12874 / 19962 =  64.493
Robust  acc:  4436 /  7535 =  58.872
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6143 /  9767 =  62.895
0, 1  acc:  4436 /  7535 =  58.872
1, 0  acc:  2121 /  2480 =  85.524
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12874 / 19962 =  64.493
Robust  acc:  4436 /  7535 =  58.872
------------------------------------
Epoch:  26 | Train Loss: 0.001 | Train Acc: 89.110 | Val Loss: 0.003 | Val Acc: 60.734
Training:
Accuracies by groups:
0, 0  acc: 12032 / 21637 =  55.608
0, 1  acc:  2683 /  5414 =  49.557
1, 0  acc: 111315 / 116535 =  95.521
1, 1  acc: 19014 / 19184 =  99.114
--------------------------------------
Average acc: 145044 / 162770 =  89.110
Robust  acc:  2683 /  5414 =  49.557
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4659 /  8535 =  54.587
0, 1  acc:  4685 /  8276 =  56.609
1, 0  acc:  2543 /  2874 =  88.483
1, 1  acc:   179 /   182 =  98.352
------------------------------------
Average acc: 12066 / 19867 =  60.734
Robust  acc:  4659 /  8535 =  54.587
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 62.369
Robust Acc: 55.926 | Best Acc: 98.333
-------------------------------------
Training, Epoch 25:
Accuracies by groups:
0, 0  acc:  5868 /  9767 =  60.080
0, 1  acc:  4214 /  7535 =  55.926
1, 0  acc:  2191 /  2480 =  88.347
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12450 / 19962 =  62.369
Robust  acc:  4214 /  7535 =  55.926
------------------------------------
Accuracies by groups:
0, 0  acc:  5868 /  9767 =  60.080
0, 1  acc:  4214 /  7535 =  55.926
1, 0  acc:  2191 /  2480 =  88.347
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12450 / 19962 =  62.369
Robust  acc:  4214 /  7535 =  55.926
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5868 /  9767 =  60.080
0, 1  acc:  4214 /  7535 =  55.926
1, 0  acc:  2191 /  2480 =  88.347
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12450 / 19962 =  62.369
Robust  acc:  4214 /  7535 =  55.926
------------------------------------
Epoch:  27 | Train Loss: 0.001 | Train Acc: 89.312 | Val Loss: 0.003 | Val Acc: 62.873
Training:
Accuracies by groups:
0, 0  acc: 12594 / 21999 =  57.248
0, 1  acc:  2697 /  5352 =  50.392
1, 0  acc: 111223 / 116373 =  95.575
1, 1  acc: 18859 / 19046 =  99.018
--------------------------------------
Average acc: 145373 / 162770 =  89.312
Robust  acc:  2697 /  5352 =  50.392
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4962 /  8535 =  58.137
0, 1  acc:  4924 /  8276 =  59.497
1, 0  acc:  2428 /  2874 =  84.482
1, 1  acc:   177 /   182 =  97.253
------------------------------------
Average acc: 12491 / 19867 =  62.873
Robust  acc:  4962 /  8535 =  58.137
------------------------------------
New max robust acc: 58.13708260105448
debias model - Saving best checkpoint at epoch 26
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=24-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=26-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.503
Robust Acc: 58.845 | Best Acc: 98.333
-------------------------------------
Training, Epoch 26:
Accuracies by groups:
0, 0  acc:  6163 /  9767 =  63.100
0, 1  acc:  4434 /  7535 =  58.845
1, 0  acc:  2102 /  2480 =  84.758
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12876 / 19962 =  64.503
Robust  acc:  4434 /  7535 =  58.845
------------------------------------
Accuracies by groups:
0, 0  acc:  6163 /  9767 =  63.100
0, 1  acc:  4434 /  7535 =  58.845
1, 0  acc:  2102 /  2480 =  84.758
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12876 / 19962 =  64.503
Robust  acc:  4434 /  7535 =  58.845
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6163 /  9767 =  63.100
0, 1  acc:  4434 /  7535 =  58.845
1, 0  acc:  2102 /  2480 =  84.758
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12876 / 19962 =  64.503
Robust  acc:  4434 /  7535 =  58.845
------------------------------------
Epoch:  28 | Train Loss: 0.001 | Train Acc: 89.708 | Val Loss: 0.003 | Val Acc: 62.123
Training:
Accuracies by groups:
0, 0  acc: 12748 / 21728 =  58.671
0, 1  acc:  2861 /  5464 =  52.361
1, 0  acc: 111516 / 116501 =  95.721
1, 1  acc: 18892 / 19077 =  99.030
--------------------------------------
Average acc: 146017 / 162770 =  89.708
Robust  acc:  2861 /  5464 =  52.361
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4940 /  8535 =  57.879
0, 1  acc:  4950 /  8276 =  59.812
1, 0  acc:  2275 /  2874 =  79.158
1, 1  acc:   177 /   182 =  97.253
------------------------------------
Average acc: 12342 / 19867 =  62.123
Robust  acc:  4940 /  8535 =  57.879
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.132
Robust Acc: 59.748 | Best Acc: 96.111
-------------------------------------
Training, Epoch 27:
Accuracies by groups:
0, 0  acc:  6174 /  9767 =  63.213
0, 1  acc:  4502 /  7535 =  59.748
1, 0  acc:  1953 /  2480 =  78.750
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 12802 / 19962 =  64.132
Robust  acc:  4502 /  7535 =  59.748
------------------------------------
Accuracies by groups:
0, 0  acc:  6174 /  9767 =  63.213
0, 1  acc:  4502 /  7535 =  59.748
1, 0  acc:  1953 /  2480 =  78.750
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 12802 / 19962 =  64.132
Robust  acc:  4502 /  7535 =  59.748
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6174 /  9767 =  63.213
0, 1  acc:  4502 /  7535 =  59.748
1, 0  acc:  1953 /  2480 =  78.750
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 12802 / 19962 =  64.132
Robust  acc:  4502 /  7535 =  59.748
------------------------------------
Epoch:  29 | Train Loss: 0.001 | Train Acc: 90.127 | Val Loss: 0.003 | Val Acc: 55.841
Training:
Accuracies by groups:
0, 0  acc: 13004 / 21505 =  60.470
0, 1  acc:  2851 /  5359 =  53.200
1, 0  acc: 112011 / 116891 =  95.825
1, 1  acc: 18834 / 19015 =  99.048
--------------------------------------
Average acc: 146700 / 162770 =  90.127
Robust  acc:  2851 /  5359 =  53.200
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4267 /  8535 =  49.994
0, 1  acc:  4322 /  8276 =  52.223
1, 0  acc:  2327 /  2874 =  80.967
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 11094 / 19867 =  55.841
Robust  acc:  4267 /  8535 =  49.994
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 57.604
Robust Acc: 51.228 | Best Acc: 95.000
-------------------------------------
Training, Epoch 28:
Accuracies by groups:
0, 0  acc:  5426 /  9767 =  55.554
0, 1  acc:  3860 /  7535 =  51.228
1, 0  acc:  2042 /  2480 =  82.339
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 11499 / 19962 =  57.604
Robust  acc:  3860 /  7535 =  51.228
------------------------------------
Accuracies by groups:
0, 0  acc:  5426 /  9767 =  55.554
0, 1  acc:  3860 /  7535 =  51.228
1, 0  acc:  2042 /  2480 =  82.339
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 11499 / 19962 =  57.604
Robust  acc:  3860 /  7535 =  51.228
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5426 /  9767 =  55.554
0, 1  acc:  3860 /  7535 =  51.228
1, 0  acc:  2042 /  2480 =  82.339
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 11499 / 19962 =  57.604
Robust  acc:  3860 /  7535 =  51.228
------------------------------------
Epoch:  30 | Train Loss: 0.001 | Train Acc: 90.163 | Val Loss: 0.004 | Val Acc: 52.328
Training:
Accuracies by groups:
0, 0  acc: 13377 / 21901 =  61.079
0, 1  acc:  2993 /  5530 =  54.123
1, 0  acc: 111413 / 116157 =  95.916
1, 1  acc: 18976 / 19182 =  98.926
--------------------------------------
Average acc: 146759 / 162770 =  90.163
Robust  acc:  2993 /  5530 =  54.123
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3915 /  8535 =  45.870
0, 1  acc:  4103 /  8276 =  49.577
1, 0  acc:  2198 /  2874 =  76.479
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 10396 / 19867 =  52.328
Robust  acc:  3915 /  8535 =  45.870
------------------------------------
-------------------------------------------
Avg Test Loss: 0.004 | Avg Test Acc: 53.897
Robust Acc: 48.321 | Best Acc: 95.000
-------------------------------------
Training, Epoch 29:
Accuracies by groups:
0, 0  acc:  5034 /  9767 =  51.541
0, 1  acc:  3641 /  7535 =  48.321
1, 0  acc:  1913 /  2480 =  77.137
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 10759 / 19962 =  53.897
Robust  acc:  3641 /  7535 =  48.321
------------------------------------
Accuracies by groups:
0, 0  acc:  5034 /  9767 =  51.541
0, 1  acc:  3641 /  7535 =  48.321
1, 0  acc:  1913 /  2480 =  77.137
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 10759 / 19962 =  53.897
Robust  acc:  3641 /  7535 =  48.321
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5034 /  9767 =  51.541
0, 1  acc:  3641 /  7535 =  48.321
1, 0  acc:  1913 /  2480 =  77.137
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 10759 / 19962 =  53.897
Robust  acc:  3641 /  7535 =  48.321
------------------------------------
Epoch:  31 | Train Loss: 0.001 | Train Acc: 90.456 | Val Loss: 0.003 | Val Acc: 62.410
Training:
Accuracies by groups:
0, 0  acc: 13583 / 21793 =  62.327
0, 1  acc:  2952 /  5427 =  54.395
1, 0  acc: 111699 / 116324 =  96.024
1, 1  acc: 19002 / 19226 =  98.835
--------------------------------------
Average acc: 147236 / 162770 =  90.456
Robust  acc:  2952 /  5427 =  54.395
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4833 /  8535 =  56.626
0, 1  acc:  4969 /  8276 =  60.041
1, 0  acc:  2420 /  2874 =  84.203
1, 1  acc:   177 /   182 =  97.253
------------------------------------
Average acc: 12399 / 19867 =  62.410
Robust  acc:  4833 /  8535 =  56.626
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.182
Robust Acc: 59.681 | Best Acc: 97.778
-------------------------------------
Training, Epoch 30:
Accuracies by groups:
0, 0  acc:  6035 /  9767 =  61.790
0, 1  acc:  4497 /  7535 =  59.681
1, 0  acc:  2104 /  2480 =  84.839
1, 1  acc:   176 /   180 =  97.778
------------------------------------
Average acc: 12812 / 19962 =  64.182
Robust  acc:  4497 /  7535 =  59.681
------------------------------------
Accuracies by groups:
0, 0  acc:  6035 /  9767 =  61.790
0, 1  acc:  4497 /  7535 =  59.681
1, 0  acc:  2104 /  2480 =  84.839
1, 1  acc:   176 /   180 =  97.778
------------------------------------
Average acc: 12812 / 19962 =  64.182
Robust  acc:  4497 /  7535 =  59.681
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6035 /  9767 =  61.790
0, 1  acc:  4497 /  7535 =  59.681
1, 0  acc:  2104 /  2480 =  84.839
1, 1  acc:   176 /   180 =  97.778
------------------------------------
Average acc: 12812 / 19962 =  64.182
Robust  acc:  4497 /  7535 =  59.681
------------------------------------
Epoch:  32 | Train Loss: 0.001 | Train Acc: 90.512 | Val Loss: 0.003 | Val Acc: 61.625
Training:
Accuracies by groups:
0, 0  acc: 13597 / 21711 =  62.627
0, 1  acc:  2945 /  5516 =  53.390
1, 0  acc: 111858 / 116380 =  96.114
1, 1  acc: 18926 / 19163 =  98.763
--------------------------------------
Average acc: 147326 / 162770 =  90.512
Robust  acc:  2945 /  5516 =  53.390
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4629 /  8535 =  54.236
0, 1  acc:  4811 /  8276 =  58.132
1, 0  acc:  2623 /  2874 =  91.267
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 12243 / 19867 =  61.625
Robust  acc:  4629 /  8535 =  54.236
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 63.240
Robust Acc: 57.784 | Best Acc: 98.889
-------------------------------------
Training, Epoch 31:
Accuracies by groups:
0, 0  acc:  5827 /  9767 =  59.660
0, 1  acc:  4354 /  7535 =  57.784
1, 0  acc:  2265 /  2480 =  91.331
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12624 / 19962 =  63.240
Robust  acc:  4354 /  7535 =  57.784
------------------------------------
Accuracies by groups:
0, 0  acc:  5827 /  9767 =  59.660
0, 1  acc:  4354 /  7535 =  57.784
1, 0  acc:  2265 /  2480 =  91.331
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12624 / 19962 =  63.240
Robust  acc:  4354 /  7535 =  57.784
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5827 /  9767 =  59.660
0, 1  acc:  4354 /  7535 =  57.784
1, 0  acc:  2265 /  2480 =  91.331
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12624 / 19962 =  63.240
Robust  acc:  4354 /  7535 =  57.784
------------------------------------
Epoch:  33 | Train Loss: 0.001 | Train Acc: 90.853 | Val Loss: 0.003 | Val Acc: 56.692
Training:
Accuracies by groups:
0, 0  acc: 13898 / 21906 =  63.444
0, 1  acc:  2924 /  5358 =  54.573
1, 0  acc: 112114 / 116321 =  96.383
1, 1  acc: 18945 / 19185 =  98.749
--------------------------------------
Average acc: 147881 / 162770 =  90.853
Robust  acc:  2924 /  5358 =  54.573
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4291 /  8535 =  50.275
0, 1  acc:  4441 /  8276 =  53.661
1, 0  acc:  2352 /  2874 =  81.837
1, 1  acc:   179 /   182 =  98.352
------------------------------------
Average acc: 11263 / 19867 =  56.692
Robust  acc:  4291 /  8535 =  50.275
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 58.451
Robust Acc: 52.926 | Best Acc: 96.111
-------------------------------------
Training, Epoch 32:
Accuracies by groups:
0, 0  acc:  5468 /  9767 =  55.984
0, 1  acc:  3988 /  7535 =  52.926
1, 0  acc:  2039 /  2480 =  82.218
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 11668 / 19962 =  58.451
Robust  acc:  3988 /  7535 =  52.926
------------------------------------
Accuracies by groups:
0, 0  acc:  5468 /  9767 =  55.984
0, 1  acc:  3988 /  7535 =  52.926
1, 0  acc:  2039 /  2480 =  82.218
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 11668 / 19962 =  58.451
Robust  acc:  3988 /  7535 =  52.926
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5468 /  9767 =  55.984
0, 1  acc:  3988 /  7535 =  52.926
1, 0  acc:  2039 /  2480 =  82.218
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 11668 / 19962 =  58.451
Robust  acc:  3988 /  7535 =  52.926
------------------------------------
Epoch:  34 | Train Loss: 0.001 | Train Acc: 90.976 | Val Loss: 0.003 | Val Acc: 65.430
Training:
Accuracies by groups:
0, 0  acc: 13836 / 21698 =  63.766
0, 1  acc:  2927 /  5419 =  54.014
1, 0  acc: 112255 / 116378 =  96.457
1, 1  acc: 19064 / 19275 =  98.905
--------------------------------------
Average acc: 148082 / 162770 =  90.976
Robust  acc:  2927 /  5419 =  54.014
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4957 /  8535 =  58.079
0, 1  acc:  5186 /  8276 =  62.663
1, 0  acc:  2678 /  2874 =  93.180
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 12999 / 19867 =  65.430
Robust  acc:  4957 /  8535 =  58.079
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 67.062
Robust Acc: 62.760 | Best Acc: 98.889
-------------------------------------
Training, Epoch 33:
Accuracies by groups:
0, 0  acc:  6177 /  9767 =  63.244
0, 1  acc:  4729 /  7535 =  62.760
1, 0  acc:  2303 /  2480 =  92.863
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13387 / 19962 =  67.062
Robust  acc:  4729 /  7535 =  62.760
------------------------------------
Accuracies by groups:
0, 0  acc:  6177 /  9767 =  63.244
0, 1  acc:  4729 /  7535 =  62.760
1, 0  acc:  2303 /  2480 =  92.863
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13387 / 19962 =  67.062
Robust  acc:  4729 /  7535 =  62.760
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6177 /  9767 =  63.244
0, 1  acc:  4729 /  7535 =  62.760
1, 0  acc:  2303 /  2480 =  92.863
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13387 / 19962 =  67.062
Robust  acc:  4729 /  7535 =  62.760
------------------------------------
Epoch:  35 | Train Loss: 0.001 | Train Acc: 91.302 | Val Loss: 0.003 | Val Acc: 64.197
Training:
Accuracies by groups:
0, 0  acc: 14258 / 21776 =  65.476
0, 1  acc:  2984 /  5482 =  54.433
1, 0  acc: 112285 / 116209 =  96.623
1, 1  acc: 19085 / 19303 =  98.871
--------------------------------------
Average acc: 148612 / 162770 =  91.302
Robust  acc:  2984 /  5482 =  54.433
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4889 /  8535 =  57.282
0, 1  acc:  5092 /  8276 =  61.527
1, 0  acc:  2595 /  2874 =  90.292
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 12754 / 19867 =  64.197
Robust  acc:  4889 /  8535 =  57.282
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 65.825
Robust Acc: 61.128 | Best Acc: 98.889
-------------------------------------
Training, Epoch 34:
Accuracies by groups:
0, 0  acc:  6120 /  9767 =  62.660
0, 1  acc:  4606 /  7535 =  61.128
1, 0  acc:  2236 /  2480 =  90.161
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13140 / 19962 =  65.825
Robust  acc:  4606 /  7535 =  61.128
------------------------------------
Accuracies by groups:
0, 0  acc:  6120 /  9767 =  62.660
0, 1  acc:  4606 /  7535 =  61.128
1, 0  acc:  2236 /  2480 =  90.161
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13140 / 19962 =  65.825
Robust  acc:  4606 /  7535 =  61.128
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6120 /  9767 =  62.660
0, 1  acc:  4606 /  7535 =  61.128
1, 0  acc:  2236 /  2480 =  90.161
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13140 / 19962 =  65.825
Robust  acc:  4606 /  7535 =  61.128
------------------------------------
Epoch:  36 | Train Loss: 0.001 | Train Acc: 91.362 | Val Loss: 0.003 | Val Acc: 68.304
Training:
Accuracies by groups:
0, 0  acc: 14285 / 21943 =  65.100
0, 1  acc:  3012 /  5403 =  55.747
1, 0  acc: 112343 / 116154 =  96.719
1, 1  acc: 19070 / 19270 =  98.962
--------------------------------------
Average acc: 148710 / 162770 =  91.362
Robust  acc:  3012 /  5403 =  55.747
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  5295 /  8535 =  62.039
0, 1  acc:  5440 /  8276 =  65.732
1, 0  acc:  2657 /  2874 =  92.450
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 13570 / 19867 =  68.304
Robust  acc:  5295 /  8535 =  62.039
------------------------------------
New max robust acc: 62.038664323374334
debias model - Saving best checkpoint at epoch 35
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=33-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=35-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 70.063
Robust Acc: 65.786 | Best Acc: 98.889
-------------------------------------
Training, Epoch 35:
Accuracies by groups:
0, 0  acc:  6567 /  9767 =  67.237
0, 1  acc:  4957 /  7535 =  65.786
1, 0  acc:  2284 /  2480 =  92.097
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13986 / 19962 =  70.063
Robust  acc:  4957 /  7535 =  65.786
------------------------------------
Accuracies by groups:
0, 0  acc:  6567 /  9767 =  67.237
0, 1  acc:  4957 /  7535 =  65.786
1, 0  acc:  2284 /  2480 =  92.097
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13986 / 19962 =  70.063
Robust  acc:  4957 /  7535 =  65.786
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6567 /  9767 =  67.237
0, 1  acc:  4957 /  7535 =  65.786
1, 0  acc:  2284 /  2480 =  92.097
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 13986 / 19962 =  70.063
Robust  acc:  4957 /  7535 =  65.786
------------------------------------
Epoch:  37 | Train Loss: 0.001 | Train Acc: 91.513 | Val Loss: 0.003 | Val Acc: 62.808
Training:
Accuracies by groups:
0, 0  acc: 14163 / 21739 =  65.150
0, 1  acc:  3032 /  5402 =  56.127
1, 0  acc: 112863 / 116552 =  96.835
1, 1  acc: 18897 / 19077 =  99.056
--------------------------------------
Average acc: 148955 / 162770 =  91.513
Robust  acc:  3032 /  5402 =  56.127
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4894 /  8535 =  57.340
0, 1  acc:  5131 /  8276 =  61.999
1, 0  acc:  2278 /  2874 =  79.262
1, 1  acc:   175 /   182 =  96.154
------------------------------------
Average acc: 12478 / 19867 =  62.808
Robust  acc:  4894 /  8535 =  57.340
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.928
Robust Acc: 62.110 | Best Acc: 95.556
-------------------------------------
Training, Epoch 36:
Accuracies by groups:
0, 0  acc:  6124 /  9767 =  62.701
0, 1  acc:  4680 /  7535 =  62.110
1, 0  acc:  1985 /  2480 =  80.040
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12961 / 19962 =  64.928
Robust  acc:  4680 /  7535 =  62.110
------------------------------------
Accuracies by groups:
0, 0  acc:  6124 /  9767 =  62.701
0, 1  acc:  4680 /  7535 =  62.110
1, 0  acc:  1985 /  2480 =  80.040
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12961 / 19962 =  64.928
Robust  acc:  4680 /  7535 =  62.110
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6124 /  9767 =  62.701
0, 1  acc:  4680 /  7535 =  62.110
1, 0  acc:  1985 /  2480 =  80.040
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12961 / 19962 =  64.928
Robust  acc:  4680 /  7535 =  62.110
------------------------------------
Epoch:  38 | Train Loss: 0.001 | Train Acc: 91.719 | Val Loss: 0.003 | Val Acc: 59.450
Training:
Accuracies by groups:
0, 0  acc: 14366 / 21716 =  66.154
0, 1  acc:  3063 /  5478 =  55.915
1, 0  acc: 112902 / 116448 =  96.955
1, 1  acc: 18960 / 19128 =  99.122
--------------------------------------
Average acc: 149291 / 162770 =  91.719
Robust  acc:  3063 /  5478 =  55.915
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4400 /  8535 =  51.552
0, 1  acc:  4615 /  8276 =  55.764
1, 0  acc:  2618 /  2874 =  91.093
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 11811 / 19867 =  59.450
Robust  acc:  4400 /  8535 =  51.552
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 60.720
Robust Acc: 55.289 | Best Acc: 98.889
-------------------------------------
Training, Epoch 37:
Accuracies by groups:
0, 0  acc:  5509 /  9767 =  56.404
0, 1  acc:  4166 /  7535 =  55.289
1, 0  acc:  2268 /  2480 =  91.452
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12121 / 19962 =  60.720
Robust  acc:  4166 /  7535 =  55.289
------------------------------------
Accuracies by groups:
0, 0  acc:  5509 /  9767 =  56.404
0, 1  acc:  4166 /  7535 =  55.289
1, 0  acc:  2268 /  2480 =  91.452
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12121 / 19962 =  60.720
Robust  acc:  4166 /  7535 =  55.289
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5509 /  9767 =  56.404
0, 1  acc:  4166 /  7535 =  55.289
1, 0  acc:  2268 /  2480 =  91.452
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12121 / 19962 =  60.720
Robust  acc:  4166 /  7535 =  55.289
------------------------------------
Epoch:  39 | Train Loss: 0.001 | Train Acc: 91.958 | Val Loss: 0.003 | Val Acc: 66.130
Training:
Accuracies by groups:
0, 0  acc: 14715 / 21907 =  67.170
0, 1  acc:  3068 /  5474 =  56.047
1, 0  acc: 112675 / 115967 =  97.161
1, 1  acc: 19222 / 19422 =  98.970
--------------------------------------
Average acc: 149680 / 162770 =  91.958
Robust  acc:  3068 /  5474 =  56.047
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  5086 /  8535 =  59.590
0, 1  acc:  5383 /  8276 =  65.043
1, 0  acc:  2491 /  2874 =  86.674
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 13138 / 19867 =  66.130
Robust  acc:  5086 /  8535 =  59.590
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 68.109
Robust Acc: 65.107 | Best Acc: 97.222
-------------------------------------
Training, Epoch 38:
Accuracies by groups:
0, 0  acc:  6359 /  9767 =  65.107
0, 1  acc:  4920 /  7535 =  65.295
1, 0  acc:  2142 /  2480 =  86.371
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 13596 / 19962 =  68.109
Robust  acc:  6359 /  9767 =  65.107
------------------------------------
Accuracies by groups:
0, 0  acc:  6359 /  9767 =  65.107
0, 1  acc:  4920 /  7535 =  65.295
1, 0  acc:  2142 /  2480 =  86.371
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 13596 / 19962 =  68.109
Robust  acc:  6359 /  9767 =  65.107
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6359 /  9767 =  65.107
0, 1  acc:  4920 /  7535 =  65.295
1, 0  acc:  2142 /  2480 =  86.371
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 13596 / 19962 =  68.109
Robust  acc:  6359 /  9767 =  65.107
------------------------------------
Epoch:  40 | Train Loss: 0.001 | Train Acc: 91.971 | Val Loss: 0.002 | Val Acc: 72.839
Training:
Accuracies by groups:
0, 0  acc: 14654 / 21920 =  66.852
0, 1  acc:  3139 /  5541 =  56.650
1, 0  acc: 113170 / 116408 =  97.218
1, 1  acc: 18738 / 18901 =  99.138
--------------------------------------
Average acc: 149701 / 162770 =  91.971
Robust  acc:  3139 /  5541 =  56.650
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  5638 /  8535 =  66.057
0, 1  acc:  5937 /  8276 =  71.738
1, 0  acc:  2720 /  2874 =  94.642
1, 1  acc:   176 /   182 =  96.703
------------------------------------
Average acc: 14471 / 19867 =  72.839
Robust  acc:  5638 /  8535 =  66.057
------------------------------------
New max robust acc: 66.05741066198009
debias model - Saving best checkpoint at epoch 39
replace: True
-> Updating checkpoint cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=35-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-wga-best-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=39-cpre=-1-cpb=-1.pt
-------------------------------------------
Avg Test Loss: 0.002 | Avg Test Acc: 74.732
Robust Acc: 71.178 | Best Acc: 97.222
-------------------------------------
Training, Epoch 39:
Accuracies by groups:
0, 0  acc:  6952 /  9767 =  71.178
0, 1  acc:  5436 /  7535 =  72.143
1, 0  acc:  2355 /  2480 =  94.960
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 14918 / 19962 =  74.732
Robust  acc:  6952 /  9767 =  71.178
------------------------------------
Accuracies by groups:
0, 0  acc:  6952 /  9767 =  71.178
0, 1  acc:  5436 /  7535 =  72.143
1, 0  acc:  2355 /  2480 =  94.960
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 14918 / 19962 =  74.732
Robust  acc:  6952 /  9767 =  71.178
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6952 /  9767 =  71.178
0, 1  acc:  5436 /  7535 =  72.143
1, 0  acc:  2355 /  2480 =  94.960
1, 1  acc:   175 /   180 =  97.222
------------------------------------
Average acc: 14918 / 19962 =  74.732
Robust  acc:  6952 /  9767 =  71.178
------------------------------------
Epoch:  41 | Train Loss: 0.001 | Train Acc: 92.116 | Val Loss: 0.003 | Val Acc: 63.538
Training:
Accuracies by groups:
0, 0  acc: 14770 / 22054 =  66.972
0, 1  acc:  3167 /  5446 =  58.153
1, 0  acc: 113162 / 116238 =  97.354
1, 1  acc: 18839 / 19032 =  98.986
--------------------------------------
Average acc: 149938 / 162770 =  92.116
Robust  acc:  3167 /  5446 =  58.153
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4898 /  8535 =  57.387
0, 1  acc:  5123 /  8276 =  61.902
1, 0  acc:  2424 /  2874 =  84.342
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 12623 / 19867 =  63.538
Robust  acc:  4898 /  8535 =  57.387
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 65.895
Robust Acc: 62.482 | Best Acc: 95.556
-------------------------------------
Training, Epoch 40:
Accuracies by groups:
0, 0  acc:  6178 /  9767 =  63.254
0, 1  acc:  4708 /  7535 =  62.482
1, 0  acc:  2096 /  2480 =  84.516
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 13154 / 19962 =  65.895
Robust  acc:  4708 /  7535 =  62.482
------------------------------------
Accuracies by groups:
0, 0  acc:  6178 /  9767 =  63.254
0, 1  acc:  4708 /  7535 =  62.482
1, 0  acc:  2096 /  2480 =  84.516
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 13154 / 19962 =  65.895
Robust  acc:  4708 /  7535 =  62.482
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6178 /  9767 =  63.254
0, 1  acc:  4708 /  7535 =  62.482
1, 0  acc:  2096 /  2480 =  84.516
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 13154 / 19962 =  65.895
Robust  acc:  4708 /  7535 =  62.482
------------------------------------
Epoch:  42 | Train Loss: 0.001 | Train Acc: 92.317 | Val Loss: 0.003 | Val Acc: 59.269
Training:
Accuracies by groups:
0, 0  acc: 14555 / 21664 =  67.185
0, 1  acc:  3128 /  5412 =  57.797
1, 0  acc: 113675 / 116594 =  97.496
1, 1  acc: 18906 / 19100 =  98.984
--------------------------------------
Average acc: 150264 / 162770 =  92.317
Robust  acc:  3128 /  5412 =  57.797
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4445 /  8535 =  52.080
0, 1  acc:  4710 /  8276 =  56.912
1, 0  acc:  2444 /  2874 =  85.038
1, 1  acc:   176 /   182 =  96.703
------------------------------------
Average acc: 11775 / 19867 =  59.269
Robust  acc:  4445 /  8535 =  52.080
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 61.016
Robust Acc: 56.297 | Best Acc: 96.667
-------------------------------------
Training, Epoch 41:
Accuracies by groups:
0, 0  acc:  5661 /  9767 =  57.960
0, 1  acc:  4242 /  7535 =  56.297
1, 0  acc:  2103 /  2480 =  84.798
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12180 / 19962 =  61.016
Robust  acc:  4242 /  7535 =  56.297
------------------------------------
Accuracies by groups:
0, 0  acc:  5661 /  9767 =  57.960
0, 1  acc:  4242 /  7535 =  56.297
1, 0  acc:  2103 /  2480 =  84.798
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12180 / 19962 =  61.016
Robust  acc:  4242 /  7535 =  56.297
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5661 /  9767 =  57.960
0, 1  acc:  4242 /  7535 =  56.297
1, 0  acc:  2103 /  2480 =  84.798
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 12180 / 19962 =  61.016
Robust  acc:  4242 /  7535 =  56.297
------------------------------------
Epoch:  43 | Train Loss: 0.001 | Train Acc: 92.242 | Val Loss: 0.003 | Val Acc: 67.217
Training:
Accuracies by groups:
0, 0  acc: 14506 / 21723 =  66.777
0, 1  acc:  3174 /  5525 =  57.448
1, 0  acc: 113802 / 116681 =  97.533
1, 1  acc: 18661 / 18841 =  99.045
--------------------------------------
Average acc: 150143 / 162770 =  92.242
Robust  acc:  3174 /  5525 =  57.448
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  5199 /  8535 =  60.914
0, 1  acc:  5487 /  8276 =  66.300
1, 0  acc:  2492 /  2874 =  86.708
1, 1  acc:   176 /   182 =  96.703
------------------------------------
Average acc: 13354 / 19867 =  67.217
Robust  acc:  5199 /  8535 =  60.914
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 69.121
Robust Acc: 65.701 | Best Acc: 96.667
-------------------------------------
Training, Epoch 42:
Accuracies by groups:
0, 0  acc:  6417 /  9767 =  65.701
0, 1  acc:  5018 /  7535 =  66.596
1, 0  acc:  2189 /  2480 =  88.266
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 13798 / 19962 =  69.121
Robust  acc:  6417 /  9767 =  65.701
------------------------------------
Accuracies by groups:
0, 0  acc:  6417 /  9767 =  65.701
0, 1  acc:  5018 /  7535 =  66.596
1, 0  acc:  2189 /  2480 =  88.266
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 13798 / 19962 =  69.121
Robust  acc:  6417 /  9767 =  65.701
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6417 /  9767 =  65.701
0, 1  acc:  5018 /  7535 =  66.596
1, 0  acc:  2189 /  2480 =  88.266
1, 1  acc:   174 /   180 =  96.667
------------------------------------
Average acc: 13798 / 19962 =  69.121
Robust  acc:  6417 /  9767 =  65.701
------------------------------------
Epoch:  44 | Train Loss: 0.001 | Train Acc: 92.301 | Val Loss: 0.003 | Val Acc: 63.049
Training:
Accuracies by groups:
0, 0  acc: 14793 / 21994 =  67.259
0, 1  acc:  3126 /  5474 =  57.106
1, 0  acc: 113324 / 116155 =  97.563
1, 1  acc: 18996 / 19147 =  99.211
--------------------------------------
Average acc: 150239 / 162770 =  92.301
Robust  acc:  3126 /  5474 =  57.106
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4721 /  8535 =  55.313
0, 1  acc:  5032 /  8276 =  60.802
1, 0  acc:  2595 /  2874 =  90.292
1, 1  acc:   178 /   182 =  97.802
------------------------------------
Average acc: 12526 / 19867 =  63.049
Robust  acc:  4721 /  8535 =  55.313
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.858
Robust Acc: 60.637 | Best Acc: 98.333
-------------------------------------
Training, Epoch 43:
Accuracies by groups:
0, 0  acc:  5950 /  9767 =  60.919
0, 1  acc:  4569 /  7535 =  60.637
1, 0  acc:  2251 /  2480 =  90.766
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12947 / 19962 =  64.858
Robust  acc:  4569 /  7535 =  60.637
------------------------------------
Accuracies by groups:
0, 0  acc:  5950 /  9767 =  60.919
0, 1  acc:  4569 /  7535 =  60.637
1, 0  acc:  2251 /  2480 =  90.766
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12947 / 19962 =  64.858
Robust  acc:  4569 /  7535 =  60.637
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5950 /  9767 =  60.919
0, 1  acc:  4569 /  7535 =  60.637
1, 0  acc:  2251 /  2480 =  90.766
1, 1  acc:   177 /   180 =  98.333
------------------------------------
Average acc: 12947 / 19962 =  64.858
Robust  acc:  4569 /  7535 =  60.637
------------------------------------
Epoch:  45 | Train Loss: 0.001 | Train Acc: 92.559 | Val Loss: 0.003 | Val Acc: 63.095
Training:
Accuracies by groups:
0, 0  acc: 14641 / 21593 =  67.804
0, 1  acc:  3185 /  5501 =  57.899
1, 0  acc: 113815 / 116489 =  97.705
1, 1  acc: 19017 / 19187 =  99.114
--------------------------------------
Average acc: 150658 / 162770 =  92.559
Robust  acc:  3185 /  5501 =  57.899
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4812 /  8535 =  56.380
0, 1  acc:  5103 /  8276 =  61.660
1, 0  acc:  2444 /  2874 =  85.038
1, 1  acc:   176 /   182 =  96.703
------------------------------------
Average acc: 12535 / 19867 =  63.095
Robust  acc:  4812 /  8535 =  56.380
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 65.284
Robust Acc: 61.792 | Best Acc: 95.000
-------------------------------------
Training, Epoch 44:
Accuracies by groups:
0, 0  acc:  6077 /  9767 =  62.220
0, 1  acc:  4656 /  7535 =  61.792
1, 0  acc:  2128 /  2480 =  85.806
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 13032 / 19962 =  65.284
Robust  acc:  4656 /  7535 =  61.792
------------------------------------
Accuracies by groups:
0, 0  acc:  6077 /  9767 =  62.220
0, 1  acc:  4656 /  7535 =  61.792
1, 0  acc:  2128 /  2480 =  85.806
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 13032 / 19962 =  65.284
Robust  acc:  4656 /  7535 =  61.792
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6077 /  9767 =  62.220
0, 1  acc:  4656 /  7535 =  61.792
1, 0  acc:  2128 /  2480 =  85.806
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 13032 / 19962 =  65.284
Robust  acc:  4656 /  7535 =  61.792
------------------------------------
Epoch:  46 | Train Loss: 0.001 | Train Acc: 92.564 | Val Loss: 0.003 | Val Acc: 61.595
Training:
Accuracies by groups:
0, 0  acc: 14753 / 21957 =  67.190
0, 1  acc:  3104 /  5288 =  58.699
1, 0  acc: 113816 / 116382 =  97.795
1, 1  acc: 18994 / 19143 =  99.222
--------------------------------------
Average acc: 150667 / 162770 =  92.564
Robust  acc:  3104 /  5288 =  58.699
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4565 /  8535 =  53.486
0, 1  acc:  4862 /  8276 =  58.748
1, 0  acc:  2631 /  2874 =  91.545
1, 1  acc:   179 /   182 =  98.352
------------------------------------
Average acc: 12237 / 19867 =  61.595
Robust  acc:  4565 /  8535 =  53.486
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 63.901
Robust Acc: 59.527 | Best Acc: 98.889
-------------------------------------
Training, Epoch 45:
Accuracies by groups:
0, 0  acc:  5814 /  9767 =  59.527
0, 1  acc:  4494 /  7535 =  59.642
1, 0  acc:  2270 /  2480 =  91.532
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12756 / 19962 =  63.901
Robust  acc:  5814 /  9767 =  59.527
------------------------------------
Accuracies by groups:
0, 0  acc:  5814 /  9767 =  59.527
0, 1  acc:  4494 /  7535 =  59.642
1, 0  acc:  2270 /  2480 =  91.532
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12756 / 19962 =  63.901
Robust  acc:  5814 /  9767 =  59.527
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5814 /  9767 =  59.527
0, 1  acc:  4494 /  7535 =  59.642
1, 0  acc:  2270 /  2480 =  91.532
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 12756 / 19962 =  63.901
Robust  acc:  5814 /  9767 =  59.527
------------------------------------
Epoch:  47 | Train Loss: 0.001 | Train Acc: 92.634 | Val Loss: 0.003 | Val Acc: 62.435
Training:
Accuracies by groups:
0, 0  acc: 14965 / 22072 =  67.801
0, 1  acc:  3201 /  5448 =  58.756
1, 0  acc: 113722 / 116193 =  97.873
1, 1  acc: 18893 / 19057 =  99.139
--------------------------------------
Average acc: 150781 / 162770 =  92.634
Robust  acc:  3201 /  5448 =  58.756
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4715 /  8535 =  55.243
0, 1  acc:  5013 /  8276 =  60.573
1, 0  acc:  2501 /  2874 =  87.022
1, 1  acc:   175 /   182 =  96.154
------------------------------------
Average acc: 12404 / 19867 =  62.435
Robust  acc:  4715 /  8535 =  55.243
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 64.072
Robust Acc: 60.332 | Best Acc: 95.000
-------------------------------------
Training, Epoch 46:
Accuracies by groups:
0, 0  acc:  5908 /  9767 =  60.489
0, 1  acc:  4546 /  7535 =  60.332
1, 0  acc:  2165 /  2480 =  87.298
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 12790 / 19962 =  64.072
Robust  acc:  4546 /  7535 =  60.332
------------------------------------
Accuracies by groups:
0, 0  acc:  5908 /  9767 =  60.489
0, 1  acc:  4546 /  7535 =  60.332
1, 0  acc:  2165 /  2480 =  87.298
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 12790 / 19962 =  64.072
Robust  acc:  4546 /  7535 =  60.332
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5908 /  9767 =  60.489
0, 1  acc:  4546 /  7535 =  60.332
1, 0  acc:  2165 /  2480 =  87.298
1, 1  acc:   171 /   180 =  95.000
------------------------------------
Average acc: 12790 / 19962 =  64.072
Robust  acc:  4546 /  7535 =  60.332
------------------------------------
Epoch:  48 | Train Loss: 0.001 | Train Acc: 92.881 | Val Loss: 0.003 | Val Acc: 59.143
Training:
Accuracies by groups:
0, 0  acc: 14768 / 21613 =  68.329
0, 1  acc:  3210 /  5424 =  59.181
1, 0  acc: 114215 / 116564 =  97.985
1, 1  acc: 18989 / 19169 =  99.061
--------------------------------------
Average acc: 151182 / 162770 =  92.881
Robust  acc:  3210 /  5424 =  59.181
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  4296 /  8535 =  50.334
0, 1  acc:  4733 /  8276 =  57.189
1, 0  acc:  2544 /  2874 =  88.518
1, 1  acc:   177 /   182 =  97.253
------------------------------------
Average acc: 11750 / 19867 =  59.143
Robust  acc:  4296 /  8535 =  50.334
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 60.916
Robust Acc: 56.466 | Best Acc: 95.556
-------------------------------------
Training, Epoch 47:
Accuracies by groups:
0, 0  acc:  5515 /  9767 =  56.466
0, 1  acc:  4281 /  7535 =  56.815
1, 0  acc:  2192 /  2480 =  88.387
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12160 / 19962 =  60.916
Robust  acc:  5515 /  9767 =  56.466
------------------------------------
Accuracies by groups:
0, 0  acc:  5515 /  9767 =  56.466
0, 1  acc:  4281 /  7535 =  56.815
1, 0  acc:  2192 /  2480 =  88.387
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12160 / 19962 =  60.916
Robust  acc:  5515 /  9767 =  56.466
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  5515 /  9767 =  56.466
0, 1  acc:  4281 /  7535 =  56.815
1, 0  acc:  2192 /  2480 =  88.387
1, 1  acc:   172 /   180 =  95.556
------------------------------------
Average acc: 12160 / 19962 =  60.916
Robust  acc:  5515 /  9767 =  56.466
------------------------------------
Epoch:  49 | Train Loss: 0.001 | Train Acc: 92.892 | Val Loss: 0.003 | Val Acc: 66.688
Training:
Accuracies by groups:
0, 0  acc: 14769 / 21580 =  68.438
0, 1  acc:  3139 /  5406 =  58.065
1, 0  acc: 114370 / 116685 =  98.016
1, 1  acc: 18922 / 19099 =  99.073
--------------------------------------
Average acc: 151200 / 162770 =  92.892
Robust  acc:  3139 /  5406 =  58.065
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  5083 /  8535 =  59.555
0, 1  acc:  5450 /  8276 =  65.853
1, 0  acc:  2540 /  2874 =  88.379
1, 1  acc:   176 /   182 =  96.703
------------------------------------
Average acc: 13249 / 19867 =  66.688
Robust  acc:  5083 /  8535 =  59.555
------------------------------------
-------------------------------------------
Avg Test Loss: 0.003 | Avg Test Acc: 68.185
Robust Acc: 64.841 | Best Acc: 96.111
-------------------------------------
Training, Epoch 48:
Accuracies by groups:
0, 0  acc:  6333 /  9767 =  64.841
0, 1  acc:  4909 /  7535 =  65.149
1, 0  acc:  2196 /  2480 =  88.548
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 13611 / 19962 =  68.185
Robust  acc:  6333 /  9767 =  64.841
------------------------------------
Accuracies by groups:
0, 0  acc:  6333 /  9767 =  64.841
0, 1  acc:  4909 /  7535 =  65.149
1, 0  acc:  2196 /  2480 =  88.548
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 13611 / 19962 =  68.185
Robust  acc:  6333 /  9767 =  64.841
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  6333 /  9767 =  64.841
0, 1  acc:  4909 /  7535 =  65.149
1, 0  acc:  2196 /  2480 =  88.548
1, 1  acc:   173 /   180 =  96.111
------------------------------------
Average acc: 13611 / 19962 =  68.185
Robust  acc:  6333 /  9767 =  64.841
------------------------------------
Epoch:  50 | Train Loss: 0.001 | Train Acc: 92.988 | Val Loss: 0.004 | Val Acc: 52.977
Training:
Accuracies by groups:
0, 0  acc: 14748 / 21571 =  68.370
0, 1  acc:  3124 /  5337 =  58.535
1, 0  acc: 114402 / 116627 =  98.092
1, 1  acc: 19082 / 19235 =  99.205
--------------------------------------
Average acc: 151356 / 162770 =  92.988
Robust  acc:  3124 /  5337 =  58.535
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  3576 /  8535 =  41.898
0, 1  acc:  4089 /  8276 =  49.408
1, 0  acc:  2680 /  2874 =  93.250
1, 1  acc:   180 /   182 =  98.901
------------------------------------
Average acc: 10525 / 19867 =  52.977
Robust  acc:  3576 /  8535 =  41.898
------------------------------------
-------------------------------------------
Avg Test Loss: 0.004 | Avg Test Acc: 53.887
Robust Acc: 47.620 | Best Acc: 98.889
-------------------------------------
Training, Epoch 49:
Accuracies by groups:
0, 0  acc:  4651 /  9767 =  47.620
0, 1  acc:  3615 /  7535 =  47.976
1, 0  acc:  2313 /  2480 =  93.266
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10757 / 19962 =  53.887
Robust  acc:  4651 /  9767 =  47.620
------------------------------------
Accuracies by groups:
0, 0  acc:  4651 /  9767 =  47.620
0, 1  acc:  3615 /  7535 =  47.976
1, 0  acc:  2313 /  2480 =  93.266
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10757 / 19962 =  53.887
Robust  acc:  4651 /  9767 =  47.620
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  4651 /  9767 =  47.620
0, 1  acc:  3615 /  7535 =  47.976
1, 0  acc:  2313 /  2480 =  93.266
1, 1  acc:   178 /   180 =  98.889
------------------------------------
Average acc: 10757 / 19962 =  53.887
Robust  acc:  4651 /  9767 =  47.620
------------------------------------
replace: True
Checkpoint saved at ./model/celebA/config/cp-debias-end-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=49-cpre=-1-cpb=-1.pt
0.556
------------------------------------
Accuracies by groups:
0, 0  acc:  8825 /  9767 =  90.355
0, 1  acc:  7116 /  7535 =  94.439
1, 0  acc:  2263 /  2480 =  91.250
1, 1  acc:   127 /   180 =  70.556
------------------------------------
Average acc: 18331 / 19962 =  91.829
Robust  acc:   127 /   180 =  70.556
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  8825 /  9767 =  90.355
0, 1  acc:  7116 /  7535 =  94.439
1, 0  acc:  2263 /  2480 =  91.250
1, 1  acc:   127 /   180 =  70.556
------------------------------------
Average acc: 18331 / 19962 =  91.829
Robust  acc:   127 /   180 =  70.556
------------------------------------
Epoch:  49 | Train Loss: 0.001 | Train Acc: 92.013 | Val Loss: 0.001 | Val Acc: 92.183
Training:
Accuracies by groups:
0, 0  acc: 32838 / 41592 =  78.953
0, 1  acc: 12642 / 14421 =  87.664
1, 0  acc: 93283 / 95381 =  97.800
1, 1  acc: 11006 / 11376 =  96.748
--------------------------------------
Average acc: 149769 / 162770 =  92.013
Robust  acc: 32838 / 41592 =  78.953
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  7639 /  8535 =  89.502
0, 1  acc:  7898 /  8276 =  95.433
1, 0  acc:  2637 /  2874 =  91.754
1, 1  acc:   140 /   182 =  76.923
------------------------------------
Average acc: 18314 / 19867 =  92.183
Robust  acc:   140 /   182 =  76.923
------------------------------------
-------------------------------------------
Avg Test Loss: 0.001 | Avg Test Acc: 92.020
Robust Acc: 70.000 | Best Acc: 94.638
-------------------------------------
Training, Epoch 48:
Accuracies by groups:
0, 0  acc:  8852 /  9767 =  90.632
0, 1  acc:  7131 /  7535 =  94.638
1, 0  acc:  2260 /  2480 =  91.129
1, 1  acc:   126 /   180 =  70.000
------------------------------------
Average acc: 18369 / 19962 =  92.020
Robust  acc:   126 /   180 =  70.000
------------------------------------
Accuracies by groups:
0, 0  acc:  8852 /  9767 =  90.632
0, 1  acc:  7131 /  7535 =  94.638
1, 0  acc:  2260 /  2480 =  91.129
1, 1  acc:   126 /   180 =  70.000
------------------------------------
Average acc: 18369 / 19962 =  92.020
Robust  acc:   126 /   180 =  70.000
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  8852 /  9767 =  90.632
0, 1  acc:  7131 /  7535 =  94.638
1, 0  acc:  2260 /  2480 =  91.129
1, 1  acc:   126 /   180 =  70.000
------------------------------------
Average acc: 18369 / 19962 =  92.020
Robust  acc:   126 /   180 =  70.000
------------------------------------
Epoch:  50 | Train Loss: 0.001 | Train Acc: 92.387 | Val Loss: 0.001 | Val Acc: 92.289
Training:
Accuracies by groups:
0, 0  acc: 33317 / 41873 =  79.567
0, 1  acc: 12877 / 14546 =  88.526
1, 0  acc: 93290 / 95136 =  98.060
1, 1  acc: 10895 / 11215 =  97.147
--------------------------------------
Average acc: 150379 / 162770 =  92.387
Robust  acc: 33317 / 41873 =  79.567
--------------------------------------
Validating:
Accuracies by groups:
0, 0  acc:  7652 /  8535 =  89.654
0, 1  acc:  7912 /  8276 =  95.602
1, 0  acc:  2631 /  2874 =  91.545
1, 1  acc:   140 /   182 =  76.923
------------------------------------
Average acc: 18335 / 19867 =  92.289
Robust  acc:   140 /   182 =  76.923
------------------------------------
-------------------------------------------
Avg Test Loss: 0.001 | Avg Test Acc: 92.165
Robust Acc: 70.556 | Best Acc: 94.784
-------------------------------------
Training, Epoch 49:
Accuracies by groups:
0, 0  acc:  8872 /  9767 =  90.836
0, 1  acc:  7142 /  7535 =  94.784
1, 0  acc:  2257 /  2480 =  91.008
1, 1  acc:   127 /   180 =  70.556
------------------------------------
Average acc: 18398 / 19962 =  92.165
Robust  acc:   127 /   180 =  70.556
------------------------------------
Accuracies by groups:
0, 0  acc:  8872 /  9767 =  90.836
0, 1  acc:  7142 /  7535 =  94.784
1, 0  acc:  2257 /  2480 =  91.008
1, 1  acc:   127 /   180 =  70.556
------------------------------------
Average acc: 18398 / 19962 =  92.165
Robust  acc:   127 /   180 =  70.556
------------------------------------
Testing:
Accuracies by groups:
0, 0  acc:  8872 /  9767 =  90.836
0, 1  acc:  7142 /  7535 =  94.784
1, 0  acc:  2257 /  2480 =  91.008
1, 1  acc:   127 /   180 =  70.556
------------------------------------
Average acc: 18398 / 19962 =  92.165
Robust  acc:   127 /   180 =  70.556
------------------------------------
replace: True
-> Updating checkpoint cp-debias-end-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=49-cpre=-1-cpb=-1.pt...
Checkpoint saved at ./model/celebA/config/cp-debias-end-cnc-celebA-sw=re-na=32-np=32-nn=32-nne=0-tsr=1-t=0.05-bf=None-cw=0.5-sud=0-me=50-bst=256-o=sgd-lr=1e-05-mo=0.9-wd=0.1-wdc=0.1-spur-me=1-bst=32-lr=0.001-mo=0.9-wd=0.0005-sts=0.2-s=0-r=0-cpe=49-cpre=-1-cpb=-1.pt
